How to Build a Dashboard That Answers Business Questions
Design measures, views and checks around one audience and decision.
Part of the Data Analytics track: modules, projects and first roles →
By the GullyMentor team · 5 min read · Published 2026-10-18 · Updated 2026-11-06 · Sources listed at the end. Not a promise about mentorship outcomes.
A useful dashboard helps one specific person answer one specific question. Start with that decision, then choose measures and charts. Starting with a chart gallery usually produces an attractive report with no clear purpose.
What steps build a good dashboard?
Define the audience and decision, sketch the answer, build and reconcile the numbers, make interaction predictable, then test the reading with a peer. Dashboards are a core output in the data analytics track. The SQL behind them is covered in SQL skills for a data analyst.
Who is the dashboard for?
For a fictional campus shop, a manager wants to understand why weekly net sales changed. Define net sales, the comparison period and whether the current week is complete.
List the dimensions you need: product, date, quantity, amount and refunds. Avoid personal customer information when aggregate reporting is enough.
A one-line brief keeps you honest: "Help the shop manager see why net sales moved last week, by product."
How do you sketch the answer?
Place the headline measure and its period first. Follow with a trend, then a breakdown that explains the change. A table may beat a chart when exact values matter.
For example, show four complete weeks and a product contribution table. Label currency and period explicitly. Do not add a percentage-change badge without a visible comparison baseline.
A rough layout on paper:
``
[ Net sales, last complete week | vs previous week ]
[ Weekly trend, 4 weeks ]
[ Product contribution to change (table) ]
``
How do you build and reconcile it?
Create transformations in a repeatable query or workbook. Keep metric definitions in a small dictionary, for example "net sales = sales amount minus refunds, completed orders only". Check source totals, dashboard totals and filtered totals against examples you calculated independently.
Test an empty period, a missing product category and a refunded order. The dashboard should tell "no records" apart from a measured value of zero.
How do you make interaction predictable?
Filters should show their current state and carry clear labels. Keep tables readable on smaller screens, provide keyboard access and use more than colour to identify a series.
A chart title should name the question or measure. "Net sales by week" is clearer than "Performance analysis".
What are the common mistakes?
- Mixing a partial week with complete weeks in the same trend.
- Showing a percentage change with no baseline.
- Defining a metric differently on two charts.
- Using colour as the only way to tell series apart.
- Filling the page with charts nobody asked for.
How do you check it answers the question?
Ask a peer who was not involved to answer three questions: What changed? Where did it change? What should be investigated next? If they cannot, remove decoration and improve the evidence. Also compare one number on the dashboard to a hand calculation from the raw data. They must match.
Do not turn an association into a causal claim. Sales falling alongside stock shortages suggests an investigation, not proof of the full cause.
What should you deliver?
Hand over the dashboard, the metric dictionary, the source reconciliation and a short findings memo. Those supporting files make the visual credible. They also give you concrete decisions to explain in an interview.
Worked example: a week where sales fell
Net sales dropped from one week to the next. The trend chart shows the fall. The contribution table shows that one product explains most of it and that its refunds rose. That points to an investigation of that product, such as stock or quality. It does not prove the cause. Put the table next to the headline so the reader sees where to look first, and add a note on what data would confirm the reason.
Which common chart choices work?
Use a line for change over time, a bar for comparing categories and a table when people need exact values. Avoid pie charts with many slices and dual axes that mislead. Keep the same colour for the same product on every chart.
What should you build next?
Pick a brief from data analytics portfolio projects and build its dashboard with this process. Free tools such as a spreadsheet or an open source BI tool are enough to practise the thinking.
Sources & further reading
Examples and practice recommendations are educational illustrations. Tool documentation and programme inclusions should be checked for your intended use.